Real-Time Path Planning for Fully Actuated Autonomous Surface Vehicles*

Abstract

This paper presents a method for real-time path planning for fully actuated autonomous surface vehicles in confined waters. The goal is to continuously generate a collision-free path from a given initial pose to a given final pose. Both the own vehicle and static obstacles are represented as convex polygons. As soon as the environment changes, or other initial or final poses are specified, a warm start is performed, in which the results of previous solutions are reused. An optimal sampling-based approach is used to explore the search space. In a cost function, the length of the path is weighted together with the distance to all obstacles. Some parts of the cost function are calculated in advance and stored in look-up tables to reduce the computation time. The result is an optimal path from an initial pose to a final pose that avoids collisions of the vehicle with static obstacles. The proposed warm start procedure is tested by real-time experiments using different scenarios.